MCP Server Documentation

What is this?

Admin note: This feature is the bridge between your business data and modern AI assistants like ChatGPT, Claude, or LeChat. It allows you to “talk” to your e-shop data using natural language to get instant answers and analysis.

Who is this for? Administrators should read this to understand how they can use AI to explore their own data. You can set this up yourself in your favorite AI app’s settings by pasting the provided URL — no coding, no API key, no developer required for basic use, though your developer can help with more advanced configurations.

Key benefits for the e-shop:

Instant Data Answers: Ask your AI “What were my top 5 selling products last week?” or “Show me all orders from New York” and get an immediate answer.
Natural Language Analysis: No need to learn complex query languages—just talk to your data as if you were talking to a human assistant.
AI-Powered Insights: Let the AI analyze your sales trends, identify low-stock items, or summarize customer behavior for you.
Modern Workflow: Bring your e-shop data directly into the AI tools you already use every day for planning and strategy.

Overview

The Model Context Protocol (MCP) server provides standardized access to your data for AI applications. It allows AI assistants like LeChat, ChatGPT, Claude, and others to query your data directly using natural language.

Protocol: JSON-RPC 2.0 | MCP Protocol 2025-06-18 | HTTP Transport | Real-time Queries

Quick Setup

For AI assistants that support MCP connectors (ChatGPT, Claude, …), this is genuinely one step: paste the Server URL below into the assistant’s connector settings. It will find its own way to sign you in and ask you to approve access — no API key, no copying client IDs, nothing to set up beforehand.

Server Endpoint

Server URL: {HOSTNAME}/api/mcp/{companyId}/mcp

Server Info: {HOSTNAME}/api/mcp/{companyId}/info

A Settings page with this URL and copy-paste instructions is also available directly at [navigate:/settings/mcp|Settings → MCP Server|plug].

Authentication Options

The default, and what the setup guides below use. The assistant discovers how to connect on its own the moment you give it the Server URL, then opens a normal sign-in + “Approve access?” screen — the same kind of prompt you’d see connecting any other app to your account. Nothing is stored in the assistant beyond the access it’s granted, and it can be revoked at any time (see below).

API Key

For tools that don’t support that automatic sign-in flow, a static API key works the same way it always has: generate one and paste it as a Bearer token.

[navigate:/apiKeys|Get API Key|key]

Build your own integration

Developers integrating directly (rather than through an off-the-shelf AI assistant) can use the OAuth endpoints directly — including self-registering a client (no manual setup, RFC 7591) and connecting with PKCE instead of a stored secret (RFC 7636). See doc/OAUTH_SETUP.md in the codebase for the full technical walkthrough with curl examples, or use a pre-registered client_id/client_secret the traditional way if you prefer.

Discovery: {HOSTNAME}/.well-known/oauth-authorization-server
Authorize: {HOSTNAME}/auth/oauth/authorize
Token exchange: {HOSTNAME}/auth/oauth/token
Revoke: {HOSTNAME}/auth/oauth/revoke

The resulting access token is used the same way as an API key — as an Authorization: Bearer <token> header on requests to the MCP endpoint.

AI Platform Setup

Connect your favorite AI assistant to your MCP server. Every platform below follows the same shape: paste the Server URL, click Connect, sign in and approve access — no API key required.

Claude

1. Open Claude settings

Go to Settings → Connectors

2. Add a connector

Click “Add connector” and paste the Server URL: {HOSTNAME}/api/mcp/{companyId}/mcp

3. Connect and approve

Click Connect, sign in if prompted, and approve access. Try asking: “What data do you have access to?” or “Show me my products”

ChatGPT [badge:Recommended|success|right]

1. Open ChatGPT settings

Go to Settings → Connectors

2. Add a custom connector

Click “Add custom connector” and paste the Server URL: {HOSTNAME}/api/mcp/{companyId}/mcp

3. Connect and approve

Click Connect, sign in if prompted, and approve access. Try asking: “Show me my products” or “What categories do I have?”

**Note:** Some ChatGPT plans only expose connectors for research/browsing rather than the full tool set. If a connector doesn't appear where expected, check that your plan supports custom connectors.

LeChat

1. Open LeChat

Go to lechat.mistral.ai and start a new conversation

2. Access connector/tool settings

In the chat interface, look for the MCP or External Tools settings

3. Add the server

Paste the Server URL: {HOSTNAME}/api/mcp/{companyId}/mcp, then sign in and approve access when prompted

**LeChat Advantage:**
LeChat by Mistral AI has excellent MCP support and can effectively use your data tools for analysis, reporting, and data exploration tasks.

Anything else that speaks MCP

Any MCP client with remote-connector support will discover and complete sign-in the same way — paste the Server URL and follow its prompts. For a client that only supports a stdio-based local bridge (no remote OAuth), see “Build your own integration” above, or fall back to a static API key.

Available Tools

AI tools exposed by the MCP server

query_data [badge:Tool|info|right]

Execute queries using the query language

Example: products(limit: 10, where: "price > 100") { id, name, price }

list_entities [badge:Tool|info|right]

List all available entities/tables

Example: No parameters required

list_{entity} [badge:Tool|info|right]

List records from specific entities

Example: limit: 10, where: "active = true", fields: "id, name"

API Examples

Direct API calls for testing and integration

Initialize Connection

curl -X POST "{HOSTNAME}/api/mcp/{companyId}/mcp" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "initialize",
    "params": {
      "protocolVersion": "2025-06-18",
      "capabilities": {
        "roots": { "listChanged": true },
        "sampling": {}
      },
      "clientInfo": {
        "name": "TestClient",
        "version": "1.0.0"
      }
    }
  }'

Query Data

curl -X POST "{HOSTNAME}/api/mcp/{companyId}/mcp" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "jsonrpc": "2.0",
    "id": 2,
    "method": "tools/call",
    "params": {
      "name": "query_data",
      "arguments": {
        "query": "products(limit: 10) { id, name, price }"
      }
    }
  }'

Learn More